[PATCH] dma-mapping: reliably inform about DMA support for IOMMU

[PATCH] dma-mapping: reliably inform about DMA support for IOMMU

[Leon Romanovsky]

From: Leon Romanovsky <leonro@nvidia.com>

If the DMA IOMMU path is going to be used, the appropriate check should
return that DMA is supported.

Fixes: b5c58b2fdc42 ("dma-mapping: direct calls for dma-iommu")
Closes: https://lore.kernel.org/all/181e06ff-35a3-434f-b505-672f430bd1cb@notapiano
Reported-by: Nícolas F. R. A. Prado <nfraprado@collabora.com> #KernelCI
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 kernel/dma/mapping.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
index 1a7de37bd643..38d7b3239dbb 100644
--- a/kernel/dma/mapping.c
+++ b/kernel/dma/mapping.c
@@ -824,6 +824,9 @@ static int dma_supported(struct device *dev, u64 mask)
 
 	if (WARN_ON(ops && use_dma_iommu(dev)))
 		return false;
+
+	if (use_dma_iommu(dev))
+		return true;
 	/*
 	 * ->dma_supported sets the bypass flag, so we must always call
 	 * into the method here unless the device is truly direct mapped.
-- 
2.46.0

Re: [PATCH] dma-mapping: reliably inform about DMA support for IOMMU

[Robin Murphy]

On 2024-09-11 11:15 am, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@nvidia.com>
> 
> If the DMA IOMMU path is going to be used, the appropriate check should
> return that DMA is supported.

Oof, indeed the significance of what iommu_dma_ops *didn't* implement is 
a subtle one...

Reviewed-by: Robin Murphy <robin.murphy@arm.com>

> Fixes: b5c58b2fdc42 ("dma-mapping: direct calls for dma-iommu")
> Closes: https://lore.kernel.org/all/181e06ff-35a3-434f-b505-672f430bd1cb@notapiano
> Reported-by: Nícolas F. R. A. Prado <nfraprado@collabora.com> #KernelCI
> Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
> ---
>   kernel/dma/mapping.c | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
> index 1a7de37bd643..38d7b3239dbb 100644
> --- a/kernel/dma/mapping.c
> +++ b/kernel/dma/mapping.c
> @@ -824,6 +824,9 @@ static int dma_supported(struct device *dev, u64 mask)
>   
>   	if (WARN_ON(ops && use_dma_iommu(dev)))
>   		return false;
> +
> +	if (use_dma_iommu(dev))
> +		return true;
>   	/*
>   	 * ->dma_supported sets the bypass flag, so we must always call
>   	 * into the method here unless the device is truly direct mapped.

Re: [PATCH] dma-mapping: reliably inform about DMA support for IOMMU

[Nícolas F. R. A. Prado]

On Wed, Sep 11, 2024 at 01:15:00PM +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@nvidia.com>
> 
> If the DMA IOMMU path is going to be used, the appropriate check should
> return that DMA is supported.
> 
> Fixes: b5c58b2fdc42 ("dma-mapping: direct calls for dma-iommu")
> Closes: https://lore.kernel.org/all/181e06ff-35a3-434f-b505-672f430bd1cb@notapiano
> Reported-by: Nícolas F. R. A. Prado <nfraprado@collabora.com> #KernelCI
> Signed-off-by: Leon Romanovsky <leonro@nvidia.com>

Thank you for the quick patch. It fixes the reported issue.

Tested-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>

Thanks,
Nícolas

Re: [PATCH] dma-mapping: reliably inform about DMA support for IOMMU

[Christoph Hellwig]

Thanks, I've applied this.  I do like the flow of dma_supported even
less with the duplicate condition now.  Can I get a quick review for
this incremental cleanup, please?

---
From d311bfe189d12a836b4add55fdb946f37f8697fa Mon Sep 17 00:00:00 2001
From: Christoph Hellwig <hch@lst.de>
Date: Thu, 12 Sep 2024 09:21:18 +0200
Subject: dma-mapping: reflow dma_supported

dma_supported has become too much spaghetti for my taste.  Reflow it to
remove the duplicate use_dma_iommu condition and make the main path more
obvious.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 kernel/dma/mapping.c | 25 ++++++++++++++-----------
 1 file changed, 14 insertions(+), 11 deletions(-)

diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
index 022d670f8cad29..cd2a97d362cd24 100644
--- a/kernel/dma/mapping.c
+++ b/kernel/dma/mapping.c
@@ -841,20 +841,23 @@ static int dma_supported(struct device *dev, u64 mask)
 {
 	const struct dma_map_ops *ops = get_dma_ops(dev);
 
-	if (WARN_ON(ops && use_dma_iommu(dev)))
-		return false;
-
-	if (use_dma_iommu(dev))
+	if (use_dma_iommu(dev)) {
+		if (WARN_ON(ops))
+			return false;
 		return true;
+	}
+
 	/*
-	 * ->dma_supported sets the bypass flag, so we must always call
-	 * into the method here unless the device is truly direct mapped.
+	 * ->dma_supported sets and clears the bypass flag, so ignore it here
+	 * and always call into the method if there is one.
 	 */
-	if (!ops)
-		return dma_direct_supported(dev, mask);
-	if (!ops->dma_supported)
-		return 1;
-	return ops->dma_supported(dev, mask);
+	if (ops) {
+		if (!ops->dma_supported)
+			return 1;
+		return ops->dma_supported(dev, mask);
+	}
+
+	return dma_direct_supported(dev, mask);
 }
 
 bool dma_pci_p2pdma_supported(struct device *dev)
-- 
2.45.2

Re: [PATCH] dma-mapping: reliably inform about DMA support for IOMMU

[Leon Romanovsky]

On Thu, Sep 12, 2024 at 09:29:54AM +0200, Christoph Hellwig wrote:
> Thanks, I've applied this.  I do like the flow of dma_supported even
> less with the duplicate condition now.  Can I get a quick review for
> this incremental cleanup, please?
> 
> ---
> From d311bfe189d12a836b4add55fdb946f37f8697fa Mon Sep 17 00:00:00 2001
> From: Christoph Hellwig <hch@lst.de>
> Date: Thu, 12 Sep 2024 09:21:18 +0200
> Subject: dma-mapping: reflow dma_supported
> 
> dma_supported has become too much spaghetti for my taste.  Reflow it to
> remove the duplicate use_dma_iommu condition and make the main path more
> obvious.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  kernel/dma/mapping.c | 25 ++++++++++++++-----------
>  1 file changed, 14 insertions(+), 11 deletions(-)

Reviewed-by: Leon Romanovsky <leon@kernel.org>

> 
> diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
> index 022d670f8cad29..cd2a97d362cd24 100644
> --- a/kernel/dma/mapping.c
> +++ b/kernel/dma/mapping.c
> @@ -841,20 +841,23 @@ static int dma_supported(struct device *dev, u64 mask)
>  {
>  	const struct dma_map_ops *ops = get_dma_ops(dev);
>  
> -	if (WARN_ON(ops && use_dma_iommu(dev)))
> -		return false;
> -
> -	if (use_dma_iommu(dev))
> +	if (use_dma_iommu(dev)) {
> +		if (WARN_ON(ops))
> +			return false;
>  		return true;

You can write it as "return !WARN_ON(ops);" if you want to be fancy, but
your version is perfectly fine too, so ignore this comment :).

> +	}
> +
>  	/*
> -	 * ->dma_supported sets the bypass flag, so we must always call
> -	 * into the method here unless the device is truly direct mapped.
> +	 * ->dma_supported sets and clears the bypass flag, so ignore it here
> +	 * and always call into the method if there is one.
>  	 */
> -	if (!ops)
> -		return dma_direct_supported(dev, mask);
> -	if (!ops->dma_supported)
> -		return 1;
> -	return ops->dma_supported(dev, mask);
> +	if (ops) {
> +		if (!ops->dma_supported)
> +			return 1;

Worth to change it to be "return true;"

> +		return ops->dma_supported(dev, mask);
> +	}
> +
> +	return dma_direct_supported(dev, mask);
>  }
>  
>  bool dma_pci_p2pdma_supported(struct device *dev)
> -- 
> 2.45.2
>